An advocacy group has urged the Trump administration to implement safety reviews for cutting-edge artificial intelligence models before they can receive government contracts. This recommendation comes as the White House addresses the national security risks posed by Anthropic’s Mythos, a model that could enable more sophisticated cyberattacks. The group emphasizes that companies developing models should undergo this vetting process to be eligible for lucrative contracts, aligning with the growing focus on AI security within the administration, which is considering an executive order on this matter. Currently, the U.S. Center for AI Standards and Innovation has voluntary agreements for safety testing with major AI developers like OpenAI and Google, but the group calls for mandatory requirements to enhance security.

xAI: xAI focuses on developing advanced AI models like Grok to advance scientific understanding and engages in U.S. government security testing. It has reached agreements to share frontier models for national security reviews. The news lists xAI in current voluntary CAISI programs facing potential mandatory requirements.
Google: Google advances frontier AI through research divisions and participates in voluntary government evaluations of its models. It has agreements for sharing unreleased models for security testing. Referenced in the news as part of existing CAISI voluntary review programs targeted for expansion.
OpenAI: OpenAI develops frontier AI models and maintains voluntary safety testing agreements with U.S. government entities. It collaborates on classified AI applications with the Pentagon. The news positions OpenAI among labs subject to proposed mandatory reviews for government contract eligibility.
Anthropic: Anthropic builds safety-oriented AI models including the Claude series and recently previewed Mythos, which raises concerns for enabling advanced cyberattacks. Its developments are under White House scrutiny for national security implications. The advocacy group highlights Anthropic’s Mythos as prompting calls for pre-release vetting.
Microsoft: Microsoft integrates AI across cloud services and partnerships while expanding safety testing collaborations with government centers. It recently announced new agreements with CAISI for AI evaluations. Included in the news among developers under voluntary agreements proposed for mandatory standards.
U.S. Government: The U.S. Government under the Trump administration manages national security, technology policy, and government procurement processes. It engages with AI developers on risk mitigation through bodies like CAISI. The news reports advocacy for screening frontier AI models before public release and tying contracts to safety reviews.
Department of Commerce: The U.S. Department of Commerce drives economic policy including AI innovation and standards enforcement. It oversees initiatives like CAISI for secure AI development. The advocacy group recommends Congress establish a permanent AI enforcement office within the department.
Americans for Responsible Innovation: Americans for Responsible Innovation is a nonprofit dedicated to public-interest AI policy advocacy on emerging technologies. It leads coalitions pushing for AI safety legislation and frameworks. In the news, the group sent a letter urging the Trump administration to mandate safety reviews for AI labs seeking government contracts.
U.S. Center for AI Standards and Innovation: The U.S. Center for AI Standards and Innovation, under NIST, facilitates industry-government collaboration on AI testing and standards development. It recently expanded voluntary security evaluations for frontier models. The news notes its role in current voluntary agreements and proposes it lead mandatory requirements.

`json
{
“Mythos Concerns”: “Anthropic’s Mythos model has triggered White House discussions due to risks of facilitating complex cyberattacks.”,
“CAISI Expansions”: “CAISI has broadened voluntary safety testing agreements with labs including Microsoft, Google, and xAI.”
}
`